Foundation Models for Time-Series Automotive Large Scale Data

Important results from the project

Yes, a large number of data parameters recorded in operational activity by different partners were collected, compiled & shared in a standardized format for further analysis by each partner according to their needs. The project has proven that the goal of sharing data can be achieved if everyone is committed. The project used Foundation Models (FMs) in industrial contexts to understand large-scale, unstructured data with the aim of discovering values that have not been seen before.

Expected long term effects

Data is today the foundation of many optimizations, in the future it´s believed that what has been done here on a very limited scale must become a standard and on a larger scale: The creation of data lakes that can be accessed to develop solutions and accelerate optimizations. As long as relevant data isn´t made accessible to many innovation speed will remain lower. The research around using increasing FMs demonstrated its potential to understand even highly unstructured large-scale data sets.

Approach and implementation

The project was based on sharing collected data between the parties in the project from a truck combination that drove in daily operation Mon-Fri. Several parties collected different types of data that needed to be put together and anonymized. This was done and shared so that research could be done. Göteborgs Universitet also researched the handling of the large amount of data.